Robust Free Power management: Dell OpenManage Power Center

OCTOBER 2013

A PRINCIPLED TECHNOLOGIES TEST REPORT Commissioned by Dell Inc.

ROBUST FREE POWER MANAGEMENT: DELL OPENMANAGE POWER CENTER

Deploying, managing, and controlling power in the data center has become increasingly

important for enterprises of all sizes. The growth of the x86 server market in the past 15 years,

the need to serve customers and end users with ever more demanding requirements, and the

explosion of big data have led to energy and power infrastructure constraints as companies,

service providers, government agencies, and other large organizations deal with the balancing

act of providing greater levels of service to their end users with a finite amount of incoming

power from the grid. The use of blades and other high-density servers compound the problem,

as they squeeze more computing power into a smaller physical footprint, which then increases

the wattage per square foot required for large data centers.

Studies show that the issue is not going away but rather growing, and that companies

should take notice. For instance, a Computer Weekly study showed that power requirements

grew globally by 63 percent from 2011 to 2012, and estimates a 17 percent increase anticipated

for 2013.1 A separate study from Gartner shows that energy-related costs are the “fastest rising

cost in the data center,” and suggests that the problem will continue to burden enterprises

unless they can find strategies to lower consumption.2

1 www.computerweekly.com/news/2240164589/Datacentre-power-demand-grew-63-in-2012-Global-datacentre-census 2 www.gartner.com/newsroom/id/1442113

http://www.principledtechnologies.com/

http://www.computerweekly.com/news/2240164589/Datacentre-power-demand-grew-63-in-2012-Global-datacentre-census

http://www.gartner.com/newsroom/id/1442113

A Principled Technologies test report 2

Robust Free Power management: Dell OpenManage Power Center

A joint study by Microsoft and the University of California, Berkeley found that

after servers, the largest contributor to the total operating cost of a data center was the

cost of provisioning power.3 Enterprises are well aware of the impact that data center

power usage has on the bottom line, business growth, and even business continuity. In a

2012 survey by Emerson Network Power, 35 percent of respondents pegged power as

the primary limiting factor for data center capacity, above cooling and square footage.4

There are pressures from multiple angles, including policy directives, grid

constraints, and most of all, budget. For example, in the case of a green policy directive,

a company may receive orders to lower power consumption by some specific

percentage over the next quarter. Other organizations must deal with grid constraints,

where plugging in additional devices could overtax their electrical infrastructure. Finally,

companies and IT departments have strict budget constraints, and face the extreme

challenge of tackling a serious issue with little financial resources. What is a company to

do?

Given these realities, what can power-constrained enterprises do?

Improved power management policies and technologies in the data center can

help enterprises solve these specific power problems while at the same time improving

both capital expenses (CAPEX) and operational expenses (OPEX). For example,

enterprises can apply policies that set power consumption limits for servers and then

scale that effort to significantly decrease power consumption across the data center,

reducing OPEX. In many cases, capping power consumption at the server level will not

result in decreased performance, as most enterprise servers use less power than the

amount typically allocated to them. Better power management can also reduce CAPEX

through better utilization of existing infrastructure through increased server density at

the rack or data center level. As each power distribution unit (PDU) kilowatt can

represent as much as $10,000 of CAPEX,5 enterprises are wise to fully use available

power capacity rather than build out unnecessarily.

In addition to improving the bottom line, power management policies and

technology can help enterprises address business continuity and mission-critical Tier 1

application availability. Enterprises may face situations in which they have no choice to

decrease power consumption in the data center, from power shortages or outages to

even running into a power ceiling, where no additional devices can be added without

removing a device. Business continuity and service-level agreements (SLAs) can also be

3 research.microsoft.com/pubs/164427/IGCC2012.pdf 4 www.emersonnetworkpower.com/documents/en-us/brands/liebert/infographics/documents/half-dc-will-run-out-of-power-or-cooling-before-space.pdf 5 www.intelcloudbuilders.com/cloud-usage-models/pdf/Intel%20PM%20article%20v2%20Chapter1-5.pdf

http://research.microsoft.com/pubs/164427/IGCC2012.pdf

http://www.emersonnetworkpower.com/documents/en-us/brands/liebert/infographics/documents/half-dc-will-run-out-of-power-or-cooling-before-space.pdf

http://www.emersonnetworkpower.com/documents/en-us/brands/liebert/infographics/documents/half-dc-will-run-out-of-power-or-cooling-before-space.pdf

http://www.intelcloudbuilders.com/cloud-usage-models/pdf/Intel%20PM%20article%20v2%20Chapter1-5.pdf



threatened by a sudden drop in available energy supply, for example as a result of a

brownout or worst-case scenarios such as the decommissioned nuclear power plants in

Japan following the March 2011 earthquake. In either case, effective server power

management can help enterprises keep running, even extending operating time on a

generator by a factor of 1.8.6

COST EFFECTIVE POWER MANAGEMENT WITH DELL OMPC Dell OpenManage Power Center is Dell’s answer to this challenge.

Administrators can use Dell OpenManage Power Center to set power limits at different

levels—data center, room, aisle, rack, or server—and set alert notifications for various

events. They can also set different power limits for different times. For example, a

server might have a lower power limit in the evenings, when the workload decreases,

and a higher limit during the day, when performance demands are greater. Such limits

can lead to increased density in data center racks as they make it possible to safely

overprovision and add extra servers to a given circuit without fear of overloading that

circuit.

In our labs at Principled Technologies, we tested two data center power

management tools—Dell OpenManage Power Center version 2.0 and JouleX Energy

Manager for Data Centers version 4.2.0. We configured a test bed with Dell

OpenManage Power Center version 2.0 managing a group of Dell PowerEdge R720

servers and JouleX Energy Manager for Data Centers, a newly acquired Cisco product,

managing an HP ProLiant DL380p Gen8 server. We used JouleX Energy Manager as an

example of a tool available for power management that provides comparable feature

sets.

While both solutions were effective at capping power, Dell OpenManage Power

Center provided a significant cost advantage compared to JouleX and its revolving

additional costs—Dell OMPC comes free with iDRAC Enterprise Edition and supports up

to 4,000 Dell servers per OMPC instance at additional no cost. In contrast, JouleX Energy

Manager for Data Centers has a starting price of $12,500 per 500 kW for the first year,

with ongoing maintenance and support costs.7 That means that with JouleX,

organizations would continue to pay high recurring costs for power management over

the life of a data center.

In addition, the cost of JouleX is based on the amount of managed device power

consumption in kW. This cost structure can be particularly frustrating to managers of a

fast-growing server farm. Each time they add a circuit, or even more servers to existing

circuits, they must consider the additional costs for the power management product. In

6 www.intel.com/content/dam/www/public/us/en/documents/case-studies/data-center-efficiency-xeon-e5-ntt-node-manager-study.pdf 7 Price accurate as of September 2013.

http://www.intel.com/content/dam/www/public/us/en/documents/case-studies/data-center-efficiency-xeon-e5-ntt-node-manager-study.pdf

http://www.intel.com/content/dam/www/public/us/en/documents/case-studies/data-center-efficiency-xeon-e5-ntt-node-manager-study.pdf



contrast, Dell OpenManage Power Center provides power management features at no

additional cost, which can help companies operate within budget and maintain

predictable operational budgets. Dell OpenManage Power Center provides the features

you need to handle a wide range of power management tasks in your Dell server farm,

and offers the added bonus of being more cost effective.

OMPC DETAILED TEST RESULTS AND SCENARIOS

We configured Dell OpenManage Power Center to communicate with the iDRAC

management port on the Dell servers. We then configured JouleX to communicate with

the iLO management port on the HP servers for a comparable configuration. JouleX

Energy Manager for Data Centers does have the ability to communicate with the

operating system to provide additional features; however, for comparison purposes we

used both servers’ out-of-band management interfaces.

Power capping

Overview

Without power management technologies in place, the power utilization of a

server typically increases as its load increases. The ability to designate an upper limit on

power consumption—a “power cap”—for a server or group of servers offers a number

of advantages and need not negatively affect performance.

One way to use power capping is by simply assigning a percentage of the

nameplate power as a cap. For example, if a server had a nameplate power of 1,200

watts and the administrator set a cap of 50 percent, the server’s power usage would

never exceed 600 watts. Administrators often use nameplate power for circuit planning,

which can cause much of the allocated power to go unused. In the above example, a 50

percent power cap based on the nameplate power would allow you to install two

servers and double your circuit/server capacity.

Dell OMPC power capping test results

To show the power capping abilities of Dell OpenManage Power Center, we ran

a workload that is common in most data centers – a database application workload

simulating e-commerce activity. We ran this workload and used approximately 80

percent CPU capacity to simulate a heavily loaded server. For details about that

workload, see Appendix C. We first ran the workload with no power cap and recorded

the orders per minute (OPM). Then we applied a 60 percent power cap—that is, 60

percent of the upper bound power. (The Dell server reports an upper bound power

reading to OpenManage Power Center, which is the maximum power the hardware

configuration could use.)



As Figure 1 shows, the number of orders per minute decreased only 0.8 percent

with the power cap in place. In business and end user terms, this shows that you can

apply a power cap with little to no performance impact for your end user, and save that

allotted power for another device. This is because when the server was uncapped, it did

not fully utilize its CPU capacity and did not need all the allotted power. Knowing that

your workload can run with a 60 percent power cap with only a minor hit to

performance allows you to allocate the unused power to other servers, so that none of

your data center power goes to waste.

Figure 1: With a 60 percent power cap in place, the Dell PowerEdge R720 achieved 99.2 percent of the performance as the same server with no power cap.

363,637 360,654

0

100,000

200,000

300,000

400,000

No power cap(Upper bound is nameplate

value of 576 Watts)

60% power cap(Upper bound is 346 Watts)

Dell PowerEdge R720 Orders per minute with and without power cap

Running a server with a power cap can also yield performance-per-watt

advantages. To show this, we used the same OLTP workload as in the above example,

but this time used 100 percent processor utilization. We first ran with no power cap and

recorded the average power for the run. Then we ran the same workload with power

caps incrementally lower than the uncapped power usage. We recorded the OPM and

calculated the performance-per-watt. We performed this test on the HP ProLiant

DL380p Gen8 as well.

Figure 2 shows the performance-per-watt results at several power cap

configurations. While the server performance does decrease as the power cap is

applied, the performance per watt increases. This shows the server running more

efficiently at lower power levels.



Dell OMPC with Dell PowerEdge R720

Power cap set (W) Average power utilization (W)

Total OPM OPM per watt

No power cap No cap 354 397,790 1,123.7

90% of peak utilization 319 318 367,218 1,154.8



Figure 2: OLTP workload results and power utilization. Higher OPM and performance/watt are better. Lower power utilization is better.

JouleX power capping with HP DL380p Gen8 test results

To demonstrate the JouleX and HP capabilities of power capping, we repeated

this test on the HP ProLiant DL380p Gen8 using JouleX. Note that HP’s out-of-band

management (iLO) implements power capping in a different manner than Dell’s iDRAC,

and it is for this reason that the OPM per watt efficiently is lower in Figure 3 below. A

Dell server capped via iDRAC will keep the server running very close to the capped

wattage, whereas an HP server capped via iLO will maintain a wattage somewhat lower

than the applied cap. Because of this design, the HP results in this report are not a direct

comparison between OpenManage Power Center and JouleX.

Figure 3 shows the performance-per-watt results at several power cap

configurations using the JouleX Energy Manager and HP ProLiant DL380p Gen8

combination.

JouleX Energy Manager with HP ProLiant DL380p Gen8

Power cap set (W) Average power utilization (W)

Total OPM OPM per watt

No power cap No cap 366 375,142 1,025.7


80% of peak utilization 293 278 275,017 990.8

70% of peak utilization 256 243 242,874 997.9

Figure 3: OLTP workload results and power utilization. Higher OPM and performance/watt are better. Lower power utilization is better.

Implementing system priority – Dell OMPC

When using power limiting, it is essential that applications across multiple

servers get the performance you need. OpenManage Power Center lets you set a

dynamic limit and set priorities at the room, aisle, or server rack levels. With a dynamic

limit, you specify a priority—high, medium, or low—for each group based on its

performance and power needs. For example, if applying a dynamic limit on a server

rack, you specify which servers are higher or lower priority. So what does this mean for

the customer or business?



Management by server groups: Companies can set power policies based

on groups of servers at the room, aisle, or rack levels.

Priority management: Using the priority features, companies can ensure

more important systems have the power they need during peak loads.

For example, a database server backing your online storefront and

fueling your business would be higher priority than an internal file

server.

Ease of use: Administrators can set a total power cap for all these

servers and let OpenManage Power Center intelligently distribute the

power based on server demand, which means each server gets the

power it needs to fulfill its role without wasting energy.

We tested the priority feature by using an e-commerce database workload to

apply a constant demand on three Dell PowerEdge R720 servers in a rack. Each server

ran at 100 percent CPU utilization and the three servers consumed approximately 1,080

total watts. We set a dynamic limit of 900 watts and gave one server high priority, one

medium priority, and one low priority. As Figure 4 shows, total power usage dropped

dramatically when we applied the limit approximately 60 seconds into the test period,

and the usage of the three servers varied according to their priority, with the high

priority server obtaining more power resources around 140 seconds.

We then stopped the workload on the medium-priority server at approximately

the 350 second mark and allowed it to focus on background tasks. Once OpenManage

Power Center identified available watts for the high- and low-priority servers, they could

consume more power up to the dynamic limit, at approximately the 450 second mark.

Figure 4: Power utilization of each R720 dynamically adjusting with respect to set server priority and power availability.

0

50

100

150

200

250

300

350

400

0 60 120 180 240 300 360 420 480 540

Po

we

r (W

atts

)

Seconds

Power utilization by priority during dynamic power cap

Low priority

Medium priority

High priority



JouleX Energy Manager for Data Centers lets you set priorities for policies and

components, but the setup is more complex, requires a higher-level engineer, and

operates via interaction with the operating system, not out of band. In contrast, Dell

OpenManage Power Center offers a free effective priority power capping option with

little required configuration.

Reliability during a power outage – Dell OMPC

It is essential that data center servers remain functional and available for users.

Both Dell OpenManage Power Center and JouleX Energy Manager for Data Centers

include a feature that reduces power following a drop in data center power or cooling or

any other emergency.

The Dell OpenManage option is called Emergency Power Reduction (EPR). This

feature gives administrators an easy way to put servers into a low power state in the

event of an emergency such as partial loss or decrease of available power. In this

situation the servers continue to run, but do not overload the circuit breaker. It

addresses emergency power issues intelligently, since the server remains functioning

during the power event and applications stay up, which is important to end users.

You can initiate EPR by selecting a device or organizational structure (rack, aisle,

room, or data center) and clicking a button; power on that device or structure reduces

immediately. Because this feature is available at all levels, from the data center down to

an individual server, administrators can select with great precision what to slow down.

To test the power reduction feature, we ran an OLTP workload using 100

percent processor utilization on the server while running on battery backup. We first ran

the server with no power limits. We then repeated the test after selecting the

emergency power reduction option.

As Figure 5 shows, the server ran 220.4 percent longer with the emergency

power reduction feature enabled. This could be the difference between keeping your

client-facing services up and running or seeing them shut down before finding an

alternate source of power.



Figure 5: With the emergency power reduction feature enabled, the Dell PowerEdge R720 lasted more than three times as long as with it disabled. 10:24

33:19

0:00

10:00

20:00

30:00

40:00

EPR off EPR on

Min

ute

s:se

con

ds

Dell PowerEdge R720 Time the server lasted with emergency power

reduction feature enabled and disabled

JouleX Energy Manager for Data Centers offers options to reduce power in the

event of emergencies, but uses different approaches. One technique is to create a policy

that changes the processor to a low performance level. However, this approach requires

setup between JouleX and the operating system instead of just the iLO management

port. We did not set this feature because we focused our evaluation on the

management port communication on both Dell and JouleX configurations. JouleX does

let you set up power reduction in the event of an emergency, but this functionality

requires a more different and more complex configuration than the dedicated EPR

option in OpenManage Power Center, and therefore was not directly comparable.

ABOUT THE PRODUCTS WE TESTED

Dell OpenManage Power Center

With Dell OpenManage Power Center, using Intel Node Manager Technology, IT

administrators get a view into various aspects of the data center, including power usage.

Power Center is an open, standards-based power management application that can read

power usage information from Dell PowerEdge servers, power distribution units,

uninterruptable power supplies, and Dell PowerEdge blade server chassis.

According to Dell, new to Dell OpenManage Power Center 2.0 from version 1

are the following features8:

Improved user interface providing a more intuitive flow with fewer

keystrokes

8 ftp://ftp.dell.com/Manuals/all-products/esuprt_software/esuprt_ent_sys_mgmt/esuprt_ent_sys_mgmt_opnmng_pwr_centr/dell-opnmang-power-center-v2.0_Release%20Notes12_en-us.pdf

ftp://ftp.dell.com/Manuals/all-products/esuprt_software/esuprt_ent_sys_mgmt/esuprt_ent_sys_mgmt_opnmng_pwr_centr/dell-opnmang-power-center-v2.0_Release Notes12_en-us.pdf

ftp://ftp.dell.com/Manuals/all-products/esuprt_software/esuprt_ent_sys_mgmt/esuprt_ent_sys_mgmt_opnmng_pwr_centr/dell-opnmang-power-center-v2.0_Release Notes12_en-us.pdf



Dashboard that offers a quick view of power headroom and device

usage

Rack power and space use data

Power monitoring and capping at the chassis level for the Dell

PowerEdge M1000e

IT administrators cannot manage power utilization unless they can collect data

related to it. OpenManage Power Center offers a range of features for monitoring your

data center power utilization. OpenManage Power Center shows peak, average, and

minimum power across the data center, room, aisle, rack, or server. Having such

granular knowledge of your power utilization can help you make improvements in many

areas, including server density, power savings, and uptime.

To learn more about Dell OpenManage Power Center, visit

www.dell.com/us/enterprise/p/dell-openmanage-power-center/pd.

JouleX Energy Manager for Data Centers

JouleX Energy Manager for Data Centers is a tool to monitor and control all

aspects of the data center. It offers support for physical and virtual servers as well as

other components such as switches, routers, storage devices and other items, but does

so at a higher software cost.

For more detailed information on JouleX Energy Manager for Data Centers, visit

www.joulex.net/products-0/joulex-energy-manager-for-data-centers/

CONCLUSION

Effective power management is quickly becoming a major concern for data

centers of all sizes. Specialized power management software is now available to help

meet these needs. We tested two of these power management tools, Dell OpenManage

Power Center and JouleX Energy Manager for Data Centers.

Both Dell OpenManage Power Center 2.0 and JouleX Energy Manager address

some of the most common critical power tasks for your data center servers. JouleX

Energy Manager is a product that includes additional features, but you must pay upfront

acquisition costs and incur recurring software maintenance costs for potentially less-

common features that just a subset of IT departments may implement. In contrast, Dell

OpenManage Power Center, when used in conjunction with iDRAC Enterprise, comes at

a very compelling price point – $0.00. This translates to a great value and an immediate

return on investment.

http://www.dell.com/us/enterprise/p/dell-openmanage-power-center/pd

http://www.joulex.net/products-0/joulex-energy-manager-for-data-centers/



APPENDIX A – SERVER CONFIGURATION INFORMATION Figure 6 provides configuration information for the two servers we tested.

System Dell PowerEdge R720 (3) HP ProLiant DL380p Gen8

Power supplies

Total number 2 2

Vendor and model number Dell 05NF18 HP HSTNS-PL29

Wattage of each (W) 750 750

Cooling fans

Total number 6 6

Vendor and model number AVC DBTC0638B2U Delta Electronics PFR0612XHE

Dimensions (h x w) of each 2.5” x 2.5” 2.5” x 2.5”

Volts 12 12

Amps 1.2 3.3

General

Number of processor packages 2 2

Number of cores per processor 8 8

Number of hardware threads per core 2 2

System power management policy Default Default

CPU

Vendor Intel® Intel

Name Xeon® Xeon

Model number E5-2680 E5-2680

Stepping C1 C1

Socket type FCLGA2011 FCLGA2011

Core frequency (GHz) 2.70 2.70

Bus frequency (GT/s) 8 GT/s 8 GT/s

L1 cache 32 KB + 32 KB (per core) 32 KB + 32 KB (per core)

L2 cache 256 KB (per core) 256 KB (per core)

L3 cache 20 MB 20 MB

Platform

Vendor and model number Dell PowerEdge R720 HP ProLiant DL380p Gen8

Motherboard model number 00W9X3 662530-001

BIOS name and version Dell 2.0.19 P70

BIOS settings Default Default

Memory module(s)

Total RAM in system (GB) 64 GB 64 GB

Vendor and model number Hynix® HMT31GR7BFR4A-H9 Samsung® M393B1K70DH0-YH9

Type PC3L-10600R PC3L-10600R

Speed (MHz) 1,333 1,333

Speed running in the system (MHz) 1,333 1,333

Timing/Latency (tCL-tRCD-tRP-tRASmin)

9-9-9-36 9-9-9-36

Size (GB) 8 8

Number of RAM modules 8 8



System Dell PowerEdge R720 (3) HP ProLiant DL380p Gen8

Chip organization Double-sided Double-sided

Rank Dual Dual

Operating system

Name Microsoft® Windows Server® 2012 Datacenter Edition

Microsoft Windows Server 2012 Datacenter Edition

Build number 9200 9200

File system NTFS NTFS

Language English English

Graphics

Vendor and model number Matrox® G200e Matrox G200e

Driver 16 MB 16 MB

RAID controller

Vendor and model number Dell PERC H710p Mini HP Smart Array P420i

Firmware version 21.1.0-0007 4.68

Driver version 6.801.5.0 (8/9/2013) 3.2.4 (8/21/2013)

Cache size 1024 MB 1024 MB

Hard drive # 1

Vendor and model number Hitachi® HUC151414CSS600 HP EH0146FAWJB

Number of disks in system 2 2

Size (GB) 146 146

Buffer size (MB) 64 16

RPM 15k 15k

Type SAS SAS

Hard drive # 2

Vendor and model number Intel SSDSC2BA20 Intel SSDSC2BA20

Number of disks in system 2 2

Size (GB) 200 200

Type SSD SSD

Ethernet adapters

Vendor and model number Intel Ethernet 10/1G 4P X540/I350 Broadcom NetExtreme Gigabit

Type Onboard Onboard

Driver 14.5.0 (9/26/2013) 16.0.0.17 (9/10/2013)

USB ports

Number 4 external, 1 internal 4 external

Type 2.0 2.0

Figure 6: System configuration information for the test servers.



APPENDIX B – CONFIGURING POWER-LIMITING SOFTWARE For Dell testing, we used Dell OpenManage Power Center version 2.0 to control the Dell PowerEdge R720 server.

OpenManage communicated with the PowerEdge R720 through the onboard iDRAC 7 Enterprise module, running

firmware version 1.46.45. We used the isolated management port for iDRAC communication.

For HP testing, we used JouleX version 4.2.0 to control the HP ProLiant DL380p Gen8. JouleX communicated with

the HP ProLiant DL380p Gen8 through the onboard iLO Advanced management module, running firmware version 1.30.

We used the isolated management port for iDRAC communication.

Installing Dell OpenManage Power Center We installed OpenManage in a virtual machine on a Dell PowerEdge R710 running Server 2012 with Hyper-V. We

configured the OpenManage Power Center VM with four vCPUs and 16 GB of memory. We used the default options for

the OpenManage installation as the installation document outlined. After installation, we used the Discovery option to

locate and configure our test servers.

Setting the power limit for Dell OpenManage Power Center 1. Select the server or rack as needed for testing. 2. Click the Policies tab. 3. Click Create New Policy to create a power limit. 4. Enter the desired limit in the Power Cap Value box. 5. Click Next through the remaining wizard windows. 6. After creating the limit, use the Enable/Disable button on the Policies tab to enable or disable the power limit. 7. Click the Emergency Power Reduction button on the device page to enable EPR for testing.

Installing JouleX Energy Manager for Data Centers We installed JouleX in a virtual machine on a Dell PowerEdge R710 running Server 2012 with Hyper-V. We

configured the JouleX VM with four vCPUs and 16 GB of memory. We used the default options for the JouleX installation

as the installation document outlined.

Setting the power limit for JouleX 1. Click the Devices tab, and click Add. 2. Enter at least an IP address or MAC address and user credentials. For our testing, we entered the IP of the HP

iLO Advanced module. Enter a name for the device in the Label field. JouleX will use this field when referring to the device.

3. Click Save Changes & Close. 4. Click the Policy tab. 5. Click Add New Rule. 6. Set a condition (e.g., the name of a device) and an action (e.g., cap power). 7. Click Save Changes.



APPENDIX C – TEST METHODOLOGY Our test setup was symmetrical between the Dell and HP configurations. On each side, we used three Dell

PowerEdge R720s (or one HP ProLiant DL380p Gen8) as our test servers. These servers ran three instances of SQL Server

2012, with two databases per instance, for a total of six databases per server under test (SUT) used in our OLTP

workload. We placed the data and log files on solid-state drives to avoid any disk bottlenecking and to stress the

processors.

Each SUT had its power run through a power meter (Yokogawa WT210s for all tests except for priority testing,

for which we added Extech 380803s). We used generic desktop client machines running Microsoft Windows Server 2003

for our power-meter software.

Each SUT had a dedicated client machine (Dell PowerEdge R720), hosting six virtual machines used as clients for

generating the OLTP workload. We used Microsoft Windows Server 2012 Datacenter Edition for all host and guest OS’s

for the SUTs and client machines.

We used two Microsoft Windows Server 2008 R2 VMs for our JouleX and Dell OpenManage Power Center

(OMPC) management services. These were placed on a Dell PowerEdge R720 running Microsoft Windows Server 2012

Datacenter.

We had two subnets in our test setup: One for workload traffic, and one for the out-of-band power

management. One onboard NIC from each SUT and client machine was dedicated to the testing network, and the

dedicated iDRAC7/iLO NIC from each SUT was used in the management network. We used a Dell PowerConnect 3548

switch to handle all traffic.

Dell IP scheme Testing network (OLTP workload) Management network (iDRAC7)

OMPC VM N/A 192.168.120.1

Client hosts 192.168.220.10/20/30 N/A

Client VMs 192.168.220.11..16/21..26/31..36 N/A

SUT1 192.168.220.1 192.168.120.11

SUT2 192.168.220.2 192.168.120.12

SUT3 192.168.220.3 192.168.120.13

HP IP scheme Testing network (OLTP workload) Management network (iDRAC7)

JouleX VM N/A 192.168.130.1

Client hosts 192.168.230.10/20/30 N/A

Client VMs 192.168.230.11..16/21..26/31..36 N/A

SUT1 192.168.230.1 192.168.130.11

Configuring the BIOS and out-of-band management

On each SUT, we set the BIOS and iDRAC/iLO to the default settings. We then accessed the iDRAC/iLO user

interfaces during boot and configured the iDRAC/iLO network settings to connect to our power management VMs.



Installing Microsoft Windows Server 2012 Datacenter Edition and SQL Server 2012

We installed Microsoft Windows Server 2012 Datacenter Edition on the SUTs, client hosts, client VMs and the

power management host. Each SUT had a pair of 146GB 15k SAS drives in a RAID1 for hosting the OS. The client hosts

had varying configurations of drives to allow space for virtual hard disks.

The SUTs also had two solid-state drives (SSDs) used for SQL Server 2012 data and log files. We created a virtual

drive from each, so that each SUT had three virtual drives in total: VD0 (146GB spindle drives in RAID1) for the OS, VD1

(single SSD) for SQL data/logs and VD2 (single SSD) for SQL data/logs.

1. Insert the installation media into the CD/DVD drive and restart the server or VM.

2. Choose the CD/DVD drive as the boot device, and boot to the installation media.

3. Press any key when prompted to boot from DVD.

4. When the installation screen appears, click My language is English.

5. Leave language, time/currency format and input method as default, and click Next.

6. Click Install now.

7. When the Windows Setup window appears, click No thanks when prompted to go online to install updates.

8. Select Windows Server 2012 Datacenter (Server with a GUI), and click Next.

9. Check I accept the license terms, and click Next.

10. Click Custom: Install Windows only (advanced).

11. Press Alt+A to open advanced partition options. Delete any partitions until there is only Drive 0 Unallocated

Space, Drive 1 Unallocated Space, and Drive 2 Unallocated Space.

12. Select Drive 0 Unallocated Space, and click Next, at which point Windows will begin installing, and will restart

automatically after completing.

13. When the Settings page appears, fill in the Password and Reenter Password fields with the same password.

14. Log in with the previously set up password.

Creating new disks in Windows Server 2012 Standard Edition

1. In the Server Manager window, click Tools, then Computer Management.

2. In the left pane, click Disk Management.

3. On the first unallocated disk (VD1), right-click the disk name and click Online. Right-click the disk name again,

and click Initialize Disk. When the Initialize Disk window pops up, select GPT, and click OK.

4. After the disk initializes, right-click the shaded area, and click New Simple Volume.

5. On the New Simple Volume Wizard that pops up, click Next.

6. Select a size (maximum, in our case), and click Next.

7. Assign a drive letter, and click Next.

8. Select Format this volume with the following settings, and leave the default settings. Click Next.

9. Click Finish.

10. Repeat for VD2.

Configuring Windows Update

1. In the left pane of the Server Manager window, click Local Server.

2. In the main frame, next to Windows Update, click Not configured.

3. In the Windows Update window, in the main pane, click Let me choose my settings.



4. Under Important updates, select Never check for updates (not recommended), and then click OK.

5. In the left pane, click Check for updates, and install all available updates.

6. Close the Windows Update window.

7. Repeat these steps to install all available updates.

Configuring Windows Firewall

1. In Server Manager, click ToolsWindows Firewall with Advanced Security.

2. In the Overview section, click Windows Firewall Properties.

3. In the Domain Profile tab, for Firewall state, click Off.

4. In the Private Profile tab, for Firewall state, click Off.

5. In the Public Profile tab, for Firewall state, click Off.

6. Then click OK.

7. Close the Windows Firewall Properties window.

Setting up Remote Desktop

1. In the Local Server tab of the Server Manager window, next to Remote Desktop, click Disabled.

2. In the System Properties window that appears, in the Remote Desktop section, select the Allow remote

connections to this computer radio button, and click OK when the warning message appears.

3. Uncheck Allow connections only from computers running Remote Desktop with Network Level Authentication

(recommended), and click OK.

Disabling IE Enhanced Security Configuration

1. In the Local Server tab of the Server Manager window, next to IE Enhanced Security Configuration, click On.

2. In the Internet Explorer Enhanced Security Configuration window, select the Off radio buttons for both

Administrators and Users, and click OK.

Installing SQL Server 2012

We installed Microsoft SQL Server 2012 on each SUT. We repeated the installation three times in order to install

three instances. We named them INSTANCE1, INSTANCE2, INSTANCE3.

1. Insert the installation DVD for SQL Server 2012 into the DVD drive.

2. Click Run SETUP.EXE. If Autoplay does not begin the installation, navigate to the SQL Server 2012 DVD, and

double-click it.

3. If the installer prompts you with a .NET installation prompt, click Yes to enable the .NET Framework Core role.

4. In the left pane, click Installation.

5. Click New SQL Server stand-alone installation or add features to an existing installation.

6. At the Setup Support Rules screen, wait for the rule check to complete. If there are no failures or relevant

warnings, click OK.

7. Enter your product key, and click Next.

8. Click the checkbox to accept the license terms, and click Next.

9. If no failures are displayed after the setup support files are installed, click Next.

10. At the Setup Role screen, choose SQL Server Feature Installation.

11. At the Feature Selection screen, select the features that your organization requires. We chose the following

features: Database Engine Services, Full-Text and Semantic Extractions for Search, Client Tools Connectivity,



Client Tools Backwards Compatibility, Management Tools – Basic, and Management Tools – Complete. Click

Next.

12. At the Installation Rules screen, click Next after the check completes.

13. At the Instance configuration screen, enter a name for the instance (INSTANCE1) and leave the default

directories.

14. At the Disk Space Requirements screen, click Next.

15. At the Server Configuration screen, leave the default users and click Next.

16. At the Database Engine Configuration screen, choose Mixed Mode and enter the system administrator

password. Click Add Current User.

17. Click the Data Directories tab, leave the default locations, and click Next. We will later manually place the OLTP

database and log files on the SSDs.

18. Click Next.

19. At the Error reporting screen, click Next.

20. At the Installation Configuration Rules screen, check that there are no failures or relevant warnings, and click

Next.

21. At the Ready to Install screen, click Install.

22. After installation completes, click Close.

23. Close the installation window.

24. Restart the server.

25. Repeat these steps for INSTANCE2 and INSTANCE3.

Configuring the Dell PowerEdge R710 client machines

We used a Dell PowerEdge R710 for each SUT to host client VMs used in our OLTP testing. We installed

Microsoft Windows Server 2012 Datacenter (according to the methodology above) on the host and enabled the Hyper-V

role.

Creating a virtual switch

1. Open Server Manager, and click ToolsHyper-V Manager.

2. In the right pane, click Virtual Switch Manager.

3. Leave External highlighted under type of virtual switch, and click Create Virtual Switch.

4. Enter a name, in our case VMSwitch, and select the NIC connected to the test network from the drop-down

menu. Check the Allow management operating system to share this network adapter, and click Apply.

5. When the warning pops up, click Yes.

6. Click OK.

Creating a virtual machine

1. In Hyper-V Manager, click New and select Virtual Machine.

2. Enter a name for the VM and click Next.

3. Enter 4096 for the Startup Memory. Click Next.

4. On the Configure Networking screen, select VMSwitch to use the virtual switch and click Next.

5. On the Connect Virtual Hard Disk screen, select the radio button for Create a virtual hard disk. Enter a name for

the VHD and a size of 40 GB. Click Finish.



6. After the VM is created, right-click the VM and click Settings.

7. Click Processor, and increase Number of virtual processors to 2.

8. Click OK.

9. Install Microsoft Windows Server 2012 Datacenter onto the VM using the methodology given above.

10. Repeat this process five more times to create six client VMs.

11. Repeat these steps on each Dell PowerEdge R710 client host.

Configuring the OLTP workload

We used the DVD Store Version 2 (DS2) benchmark to simulate our OLTP workload. DS2 is an open-source

application that models an online DVD Store where customers long in, browse movies, and purchase movies. DS2

reports its results in order per minute that the server can handle. For more details about the DS2 tool, see

http://www.delltechcenter.com/page/DVD+Store.

We ran DS2 with a 10GB database, using 64 threads and 0 think time per client. We set a warm-up time of 5

minutes to cache the database in memory, and then ran the test for 20 minutes while collecting data.

Data generation overview

We generated the data using the Install.pl script included with DVD Store version 2.1 (DS2), providing the

parameters for our 10GB database size and the database platform on which we ran: Microsoft SQL Server. We ran the

Install.pl script on a utility system running Linux. The Install.pl script also generated the database schema.

After processing the data generation, we transferred the data files and schema creation files to a Windows-

based system running SQL Server 2012. We built the 10GB database in SQL Server 2012, and then performed a full

backup, storing the backup file on the C: drive for quick access. We used that backup file to restore to the servers

between test runs. We performed the data generation procedure once.

The only modification we made to the schema creation scripts were the specified file sizes for our database. We

explicitly set the file sizes higher than necessary to ensure that no file-growth activity would affect the outputs of the

test. Besides this file size modification, the database schema was created and loaded according to the DVD Store

documentation. Specifically, we followed the steps below:

1. We generated the data and created the database and file structure using database creation scripts in the DS2

download. We made size modifications specific to our 10GB database and the appropriate changes to drive

letters.

2. We transferred the files from our Linux data generation system to a Windows system running SQL

Server.

3. We created database tables, stored procedures, and objects using the provided DVD Store scripts.

4. We set the database recovery model to bulk-logged to prevent excess logging.

5. We loaded the data we generated into the database. For data loading, we used the import wizard in SQL Server

Management Studio. Where necessary, we retained options from the original scripts, such as Enable Identity

Insert.

6. We created indices, full-text catalogs, primary keys, and foreign keys using the database-creation scripts.

7. We updated statistics on each table according to database-creation scripts, which sample 18 percent of the

table data.

http://www.delltechcenter.com/page/DVD+Store



8. On the SQL Server instance, we created a ds2user SQL Server login using the following Transact SQL (TSQL)

script:

USE [master]

GO

CREATE LOGIN [ds2user] WITH PASSWORD=N’’,

DEFAULT_DATABASE=[master],

DEFAULT_LANGUAGE=[us_english],

CHECK_EXPIRATION=OFF,

CHECK_POLICY=OFF

GO

9. We set the database recovery model back to full.

10. We created the necessary full text index using SQL Server Management Studio.

11. We created a database user and mapped this user to the SQL Server login.

We then performed a full backup of the database. This backup allowed us to restore the databases to a pristine

state relatively quickly between tests.

When restoring the database to the SUTs, we restored twice to each instance, for a total six databases per SUT.

We arranged each pair of databases to swap data and log directories between drives, as follows:

INSTANCE1

o DS2_1-1: Data on E: and Logs on F:

o DS2_1-2: Logs on E: and Data on F:

INSTANCE2



INSTANCE3



Running the DVD Store tests

We created a series of batch files, SQL scripts, and shell scripts to automate the complete test cycle. Each

complete test cycle consisted of the general steps listed below. For the 80 percent CPU test scenarios, we ran the

workload on INSTANCE1 and INSTANCE2 only. For the 100 percent CPU test scenarios, we ran the workload on all three

instances.

1. Clean up prior outputs from the server and all client driver systems.

2. Drop all databases from the test servers.

3. Restore all databases on all test servers.

4. Reboot the server and all client systems.

5. Let the test server idle until the power utilization settled.

6. Start the DVD Store driver on all respective clients.



7. We used the following DVD Store parameters for testing the servers in this study:

ds2sqlserverdriver.exe --target=<target_IP> --ramp_rate=10 –warmup_time=5 --

run_time=20 --n_threads=64 --db_size=10GB --think_time0 --detailed_view=Y –-

database_name=DS2



APPENDIX D – MEASURING POWER Measuring workload power

To record each server’s power consumption during the workload test, we used either a Yokogawa WT210 or an

Extech Instruments® (www.extech.com) 380803 Power Analyzer/Datalogger. We connected the power cord from the

server under test to the power-meter’s output load power outlet. We then plugged the power cord from the power-

meter’s input voltage connection into a 120-volt power outlet.

We used WTViewer 8.11 or the Power Analyzer’s Data Acquisition Software (version 2.11) to capture all

recordings. We installed the software on a separate Intel processor-based PC, which we connected to the power-meter

via an RS-232 cable. We captured power consumption at one-second intervals.

http://www.extech.com/



ABOUT PRINCIPLED TECHNOLOGIES

Principled Technologies, Inc. 1007 Slater Road, Suite 300 Durham, NC, 27703 www.principledtechnologies.com

We provide industry-leading technology assessment and fact-based marketing services. We bring to every assignment extensive experience with and expertise in all aspects of technology testing and analysis, from researching new technologies, to developing new methodologies, to testing with existing and new tools. When the assessment is complete, we know how to present the results to a broad range of target audiences. We provide our clients with the materials they need, from market-focused data to use in their own collateral to custom sales aids, such as test reports, performance assessments, and white papers. Every document reflects the results of our trusted independent analysis. We provide customized services that focus on our clients’ individual requirements. Whether the technology involves hardware, software, Web sites, or services, we offer the experience, expertise, and tools to help our clients assess how it will fare against its competition, its performance, its market readiness, and its quality and reliability. Our founders, Mark L. Van Name and Bill Catchings, have worked together in technology assessment for over 20 years. As journalists, they published over a thousand articles on a wide array of technology subjects. They created and led the Ziff-Davis Benchmark Operation, which developed such industry-standard benchmarks as Ziff Davis Media’s Winstone and WebBench. They founded and led eTesting Labs, and after the acquisition of that company by Lionbridge Technologies were the head and CTO of VeriTest.

Principled Technologies is a registered trademark of Principled Technologies, Inc. All other product names are the trademarks of their respective owners.

Disclaimer of Warranties; Limitation of Liability: PRINCIPLED TECHNOLOGIES, INC. HAS MADE REASONABLE EFFORTS TO ENSURE THE ACCURACY AND VALIDITY OF ITS TESTING, HOWEVER, PRINCIPLED TECHNOLOGIES, INC. SPECIFICALLY DISCLAIMS ANY WARRANTY, EXPRESSED OR IMPLIED, RELATING TO THE TEST RESULTS AND ANALYSIS, THEIR ACCURACY, COMPLETENESS OR QUALITY, INCLUDING ANY IMPLIED WARRANTY OF FITNESS FOR ANY PARTICULAR PURPOSE. ALL PERSONS OR ENTITIES RELYING ON THE RESULTS OF ANY TESTING DO SO AT THEIR OWN RISK, AND AGREE THAT PRINCIPLED TECHNOLOGIES, INC., ITS EMPLOYEES AND ITS SUBCONTRACTORS SHALL HAVE NO LIABILITY WHATSOEVER FROM ANY CLAIM OF LOSS OR DAMAGE ON ACCOUNT OF ANY ALLEGED ERROR OR DEFECT IN ANY TESTING PROCEDURE OR RESULT. IN NO EVENT SHALL PRINCIPLED TECHNOLOGIES, INC. BE LIABLE FOR INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES IN CONNECTION WITH ITS TESTING, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGES. IN NO EVENT SHALL PRINCIPLED TECHNOLOGIES, INC.’S LIABILITY, INCLUDING FOR DIRECT DAMAGES, EXCEED THE AMOUNTS PAID IN CONNECTION WITH PRINCIPLED TECHNOLOGIES, INC.’S TESTING. CUSTOMER’S SOLE AND EXCLUSIVE REMEDIES ARE AS SET FORTH HEREIN.

http://www.principledtechnologies.com/

Robust Free Power management: Dell OpenManage Power Center

Technology

Transcript of Robust Free Power management: Dell OpenManage Power Center